Electronic Letters to:
|
|
E-letters: Electronic letters published:
-
![[Read eLetter]](/icons/misc/sectionDOWN.gif)
RCT design is usually not the best option in diagnostic research
- Frans H Rutten, Johannes C Kelder, Arno W Hoes (13 February 2009)
-
Review missed the most clinically relevant diagnostic research of all.
- Ben d Ewald (10 February 2009)
-
BNP example misses the most clinically relevant points.
- Ben Ewald, Daniel Ewald (8 December 2008)
|
|
|||
|
Frans H Rutten, General Practitioner and researcher Julius Center for Health Sciences and Primary Care, UMC Utrecht, PO Box 85500, 3508GA Utrecht, Johannes C Kelder, Arno W Hoes
Send e-letter to journal:
|
Response to e-letter 'BNP example misses the most clinically relevant points'By Ben Ewald, Daniel Ewald Ewald and Ewald argue that a randomized trial is superior to a cross- sectional study in which multivariable analyses is applied when assessing the added value of a diagnostic test. We do not agree. Diagnosing in daily practice implies estimating absolute risks of a disease in suspected patients and when the aim is to assess the added value of a diagnostic test, it should be shown that the availability of the test improves these estimations. A multivariable approach in which patient characteristics, symptoms and signs are always the starting point, and the improvement in diagnostic accuracy following additional testing is quantified provides this evidence. It first shows the diagnostic accuracy of readily available items (including signs and symptoms), then the added value of the new test. Additionally, it often also allows a physician to estimate the probability of the disease for an individual patient based on his or her characteristics and findings.(1) A randomised trial (as the one performed by Wright et al)(2) comparing those patients in whom the doctor does or does not receive the results of an additional test, and the proportion correctly classified patients is compared between the two groups can also provide useful results, but such an approach has several important limitations. First, the power of such studies is considerably lower than a multivariable cross-sectional sectional study. Furthermore no information on the value of (individual or combinations of) signs and symptoms is provided and no algorithm to estimate individual probabilities can be derived. Moreover, often no standardised data on readily available items (including signs and symptoms) are assessed, and thus the true added value of the new test is not quantified. Randomised trials can be helpful in diagnostic research, for example when no excepted reference standard for the disease exists (3), but are not the best option in the assessment of B-type natriuretic peptide in suspected heart failure. (1)Rutten FH, Moons KG, Cramer MJ, Grobbee DE, Zuithoff NP, Lammers JW, Hoes AW. Recognising heart failure in elderly patients with stable chronic obstructive pulmonary disease in primary care: cross sectional diagnostic study. BMJ 2005; 331(7529):1379. (2)Wright SP, Doughty RN, Pearl A, Gamble GD, Whalley GA, Walsh HJ et al. Plasma amino-terminal pro-brain natriuretic peptide and accuracy of heart- failure diagnosis in primary care: a randomized, controlled trial. J Am Coll Cardiol 2003; 42(10):1793-1800. (3)Grobbee DE, Hoes AW. Clinical Epidemiology. Principles, Methods, and Applications for Clinical Research. 2008. Jones and Bartlett, Sudbury, USA. Conflict of Interest:None declared |
|||
|
|
|||
|
Ben d Ewald, lecturer in epidemiiology University of Newcastle NSW 2308
Send e-letter to journal:
|
Review missed the most clinically relevant diagnostic research of all. Dr Ben Ewald, University of Newcastle NSW. Kelder et al1 have pointed out some important theoretical points about the application of diagnostic test research to primary care, but have missed one of the most important as indicated in our review of this topic 2. The calls for diagnostic research to be carried out in the primary care setting and to recognise that diagnostic testing is done in an additive manner are valid, but multivariate analysis can model this process to only a limited extent. Kelder et al have not mentioned that the best evidence of all would come from a randomised controlled trial of availability of the test. This is exactly what was done by Wright et al3 in New Zealand (Kelder’s reference 31) in a study that clearly demonstrated the added information contributed by NTBNP beyond history, examination, ECG and routine blood tests. The principal benefit of the natriuretic peptide was in ruling out HF in those who did not have it. The conclusion drawn by Kelder can only be reached by ignoring the work published by Wright. references 1. Kelder JC, Rutten FH, Hoes AW. Clinically relevant diagnostic research in primary care: the example of B-type natriuretic peptides in the detection of heart failure 10.1093/fampra/cmn096. Fam. Pract. 2009;26(1):69-74. 2. Ewald B, Ewald D, Thakkinstian A, Attia J. Meta-analysis of BNP and NT- proBNP in the diagnosis of clinical heart failure and population screening for LVSD. Internal Medicine Journal 2008;38:101-113. 3. Wright SP, Doughty RN, Pearl A, Gamble GD, Whalley GA, Walsh HJ, et al. Plasma amino-terminal pro-brain natriuretic peptide and accuracy of heart- failure diagnosis in primary care*1: A randomized, controlled trial. Journal of the American College of Cardiology 2003;42(10):1793-1800. Conflict of Interest:None declared |
|||
|
|
|||
|
Ben Ewald, general practitioner newcastle NSW Australia 2300, Daniel Ewald
Send e-letter to journal:
|
Letter to the journal Family Practice Dr Ben Ewald, Dr Daniel Ewald, General Practitioners, Australia. The call by Kelder et al1 for diagnostic research in primary care to take a multivariate analysis approach does not reflect the reality of diagnosis in primary care. The paper also fails to point out the single best research design for a new diagnostic test, being a randomised controlled trial. The influence of patient variables on the interpretation of BNP values can be corrected for age by having an age specific diagnostic cut point, but cannot be corrected for the multitude of other clinical information available. What is the diagnostic cut point for someone with a normal chest Xray? What is the diagnostic cut point for someone with a prior history of AMI, or of swollen ankles? Using a complex multivariate model in the quantitative analysis of research data is of no help to the clinician who generally does not have a quantitative value of pre test probability. Even if a computer assisted decision tool were compiled based on the findings of logistic regression, it would not account for all the occurrence of heart failure, as the scope of potential predictive factors is so vast. The effect of age adjusted cut points for BNP research was examined in our meta analysis 2 and in the studies that used this approach it did not seem to increase diagnostic accuracy. Introduction of a new diagnostic test can be viewed as a health care intervention and like any other intervention can be tested with an RCT design. This has been done twice for natriuretic peptides, once in a hospital setting and once in primary care. The primary care study done in New Zealand by Wright3 randomised GPs to receive or not receive the results of NT-BNP testing, and asked them to revise their diagnosis in the light of all available information at a follow up visit. All patients had diagnostic testing (including echocardiography) and were assessed by a cardiologist as the criterion standard. This study clearly demonstrated the extra diagnostic value of NT-BNP beyond other information available to GPs, and that the greatest use was in ruling out heart failure. The strength of the design of this research makes this evidence superior to the three other primary care based studies referenced by Kelder, and answers all the problems pointed out in their theoretical discussion. A major limiting factor in the primary care application of BNP testing is that there is a lack of evidence based care for heart failure with preserved LVEF. This leaves the value of the test as primarily being to rule out LVF, as other cases will require an echocardiogram anyhow. The basic epidemiologic theory that a randomised controlled trial trumps observational research applies to diagnostic questions as it does to treatments. This fact is ignored at our peril, as demonstrated by the history of HRT which appeared highly effective and safe in observational studies. References 1) Kelder J, Rutten F, Hoes A. Clinically relevant diagnostic research in primary care: the example of B-type natriuretic peptides in the detection of heart failure. Family Practice 2008; advanced access published 3 Dec. 2) Ewald B, Ewald D, Thakkinstian A, Attia J, Meta analysis of B type natriueretic peptide and N-terminal pro BNP in the diagnosis of clinical heart failure and population screening for left ventricular systolic dysfunction. Internal Medicine Journal 2008; 38: 101-113 3) Wright S, Doughty R, Pearl A et al, Plasma amino-terminal pro- brain natriuretic peptide and accuracy of heart failure diagnosis in primary care. A randomised controlled trial. Journal of the American College of Cardiology 2003; 42: 1793-800 Conflict of Interest:None declared |
|||